Azure OpenAI Reviews

Our team has developed virtual assistants for healthcare organizations, also published in Azure Marketplace. This can be used for a personal assistant perspective. We have also developed an application for one of the fertilizer companies. Here, a farmer can go to their application, click a photo of any disease or progress in the plant, and it will identify what type of fungus or disease that plant has. Accordingly, it will recommend what kind of fertilizers and how to use them. These are a couple of use cases we have worked on.

It is easy to integrate and develop a solution. Most customers are concerned about the security of their data and how cost-effective it is. We have developed some methodologies so that our customers will not be charged too much for these OpenAI services but will still get the same kind of performance and results. It's all developed on Azure, so customers also see its benefit.

I faced one issue with Azure OpenAI: My customer wanted more clarity on the pricing. They were not able to get proper answers from the documentation or the pricing calculator. I suggest that Microsoft maintain standardization in the pricing details published in the documentation and the pricing calculator. 

Sometimes, customers check the prices independently, but the details of the pricing parameter and the documentation sheet are inconsistent. This confuses customers, making them unsure if the service is live or how to test it.

I have been working with the product for a couple of days. 

I haven't faced any issues with the tool's stability. 

The tool is highly scalable. I don't think any of my customers have faced any issues from a performance or scalability perspective when the load on their website has increased. We designed the solution to be highly scalable and reliable so they don't face challenges at the application or performance level. I have not seen any customers complaining about issues or performance problems. 

My company has seven to eight clients using Azure OpenAI. Because of data concerns, all features should be available in at least one data center in each region. For example, there are three data centers in India, and until recently, Azure OpenAI was available in only four regions. Now, it is also available in one of the data centers in India. If Microsoft introduces a new product, it should ensure it is available in at least one data center per region so that customers from that region can validate it.

Azure OpenAI's deployment is easy. It depends on a couple of things, including the solution we are developing. If it is a normal chatbot, it shouldn't take more than two to three days to complete the project. However, it might take longer if the customer has a very complex environment.

You can enjoy cost savings because you need only less expertise. Traditionally, IT services or SLM models require a lot of computing power to train models, but Microsoft has already invested in them. Customers can directly use this high-efficiency model available on the Microsoft platform. Microsoft has already done that, they don't have to build anything from scratch or use high computing resources. So, there is a good ROI.

I rate the overall product an eight out of ten. You should start by researching and experimenting with Azure OpenAI. You can create a simple chatbot or a computer vision model that identifies objects in images. This will help you understand how it works and develop use cases based on your requirements. I would also suggest testing those use cases.

In India, I'm seeing that customers across various industries, whether manufacturing, BFSI, or healthcare, are all starting to use AI in some form. Each customer develops solutions for cross-selling and processing products based on specific use cases. They design user stories for their customers, create offers to grab attention, integrate interactive chatbots into their applications or websites, and develop AI-based visitor and document management systems. Everyone wants to use AI differently, but no two use cases are the same.

Our clients are interested in building knowledge bases, particularly in child welfare. In this domain, we focus on supporting caseworkers by compiling and organizing relevant information. This information is then stored in a database using a query. The database generates summaries and reminders for specific actions and even facilitates sending emails to parents or other relevant parties.

The system's complexity is tailored to the specific needs of child welfare cases. Additionally, we're exploring opportunities to assist a healthcare organization. Specifically, we're working on streamlining the process of filling out forms required for insurance claims. This effort aims to ensure that hospitals can receive funding or payment for the care they provide.

The solution allows you to work through the extraction and summarization of unstructured documents.

The solution needs to accommodate smaller companies. 

I have been using the product for four months. 

I rate Azure OpenAI a nine out of ten. 

I rate the product's scalability a three out of ten. 

Azure OpenAI's deployment is straightforward. It is quick to deploy and can be completed in weeks. We had three resources deploying it. 

The tool costs around 20 dollars a month. 

I rate Azure OpenAI a nine out of ten. 

I use Azure OpenAI to create message dashboards for my company.

The most valuable feature of Azure OpenAI stems from the GPT-3.5 models it provides to its users. The information that I provide is based on the information I found through research. I have not started to use the solution for the development part.

The fine-tuning of models with the use of Azure OpenAI is an area with certain shortcomings currently, and it can be considered for improvement in the future. It would be great if Azure OpenAI could increase the limit of the knowledge of its chatbot. If I have datasets for fine-tuning, the chatbot can only answer the user's queries related to training data, meaning the chatbot would not provide any knowledge to its users from the outside world.

I have been using Azure OpenAI for maybe a month.

I work in a company involved in some business related to messaging, where I did not have a reason to think about scalability options provided by the solution.

I have not contacted the solution's technical support yet since I am still involved in some research related to the product.

I have experience with LangChain. LangChain uses OpenAI.

The documentation provided by Azure OpenAI is clear enough. Based on my research, I think the product's initial setup phase would be easy.

The solution is deployed on the cloud services from Azure.

Though in my company, we are in the process related to the deployment of Azure OpenAI, it has not been deployed yet.

I want Azure OpenAI since it is the best solution related to language processing.

The only suggestion I can provide to those who plan to use Azure OpenAI is to consider providing the tool with clear prompts. Azure OpenAI is one of the best tools and easy to use.

Sometimes, when I try some prompts, Azure OpenAI fails to understand them. Overall, Azure OpenAI is a really helpful tool.

I rate the overall solution a nine out of ten.

We've created a platform to build business use cases that we sell to our customers. We have partnerships with Microsoft, SAP, Databricks, and Azure OpenAI. 

We have two main focuses with Azure OpenAI:

• Enterprise Solutions: We've built solutions that act as an "enterprise copilot," digitizing and transforming processes within departments like HR, legal, finance, recruitment, IT support, and contact centers.
• Industry-Specific Solutions: We've developed solutions tailored to specific industry problems. Examples include automating airline ticket refunds, providing industry-specific customer support, helping customers plan their travel journeys, and assisting with hotel bookings.

Additionally, we integrate Azure OpenAI with data, analytics, and other areas to provide customers with 360 insights and transform customer journeys across various enterprise processes. We leverage the captured data to drive these initiatives.

OpenAI integrates seamlessly with the broader Microsoft Azure ecosystem, and that provides synergies with the other solutions. This integration makes it much easier to build solutions. 

Additionally, we can create custom solutions using Microsoft Azure development tools as needed. Since we're a software engineering organization, we can leverage the OpenAI APIs to go beyond the standard capabilities.

The modular design is very clean. The modules are very well-trained. It significantly reduces computational time compared to previous technologies. Plus, OpenAI provides a lot of training materials that allow us to quickly deploy solutions. 

There are development studios and configurations where people ensure that security is taken care of. So, Azure configurations ensure security and compliance – those guardrails are in place. All these features work seamlessly when you use Azure and OpenAI together. The security features are provided by Azure.

It also integrates with our existing enterprise tools like the broader Microsoft software suite including Office 365.

The product features themselves are fine. However, with Microsoft scaling the service so much, the support structure needs to keep pace. 

When solving complex issues, the process of interacting with Microsoft can be quite time-consuming, especially if you don't have a preferred agreement. For enterprise-level solutions that are mission-critical, real-time support is vital.

So, real-time support from Microsoft is an area for improvement, especially for complex business use cases. Right now, the support is suitable for use cases that won't directly impact the enterprise in a major way if there's a temporary issue.

Similar to what Google is doing with its marketing module, having seamless interaction with multimodal data – like videos, text, and other sources – at an optimal cost would be great. So, seamless interaction across multiple data types at optimal cost.

Multimodal interactions with optimal pricing. As use cases become more complex, and with the rise of edge devices, this becomes increasingly important.

I have been using it for 14 months now. We started using it three to four months after its initial release. 

I would rate the stability a nine out of ten. 

For smaller organizations with individual or limited use cases, the transaction costs can be a bit higher and costly. However, when we have larger-scale deployments across multiple use cases, the ability to scale is quite good. That's when the ROI (return on investment) really comes into play for an organization.

I would rate the scalability a nine out of ten. We successfully deploy it for multi-billion dollar enterprises. However, with the right configuration, it can be commercially viable for smaller businesses as well. We need to tune it properly for that use case.

Everyone has rushed into this technology. Therefore, the rapid adoption of this new technology means there are more people working with it than there are experts to provide support.

Neutral

It's almost too easy to set up. Because it's so easy, many people are jumping in and creating half-baked solutions without fully understanding the risks, security concerns, and potential problems. What's happening is that people without deep experience in operating large technology solutions are offering these solutions to businesses.

This can cause issues. For example, a recent case involved an airline in North America facing legal issues due to incorrect information provided by a chatbot. Now, there are questions of liability – is the chatbot responsible, or the business system itself? The business ended up paying a penalty.

So, responsible AI is key. The ease of setup is good from a technical standpoint, but we need governance, compliance, and checks and balances. Platform providers aren't emphasizing these aspects, and currently, there aren't regulations mandating it. So, while easy configuration is beneficial, we also need easy ways to ensure compliance, governance, security, and safeguards. All of these need to be considered – not just the ability to build something.

Since Azure itself is a cloud platform, we haven't come across many use cases where someone creates a purely private instance.

We can see results within weeks after implementation. The implementation itself might take six to ten weeks. Once live, we can see measurable differences within three to four weeks. Of course, with any new technology solution, there's a stabilization period, and the full impact is best understood over time. 

However, we can see immediate productivity benefits. Then, within six months or so, we might start seeing cost benefits. Within a year, we can potentially see improvements in feedback and metrics like NPS (Net Promoter Score).

Since everyone is on the Azure platform, Azure OpenAI's interactions with the technology generate quite a lot of data. This data can be used to quickly establish relevant digital metrics.

We've been a long-term Microsoft shop with an enterprise agreement, so that gives us some advantages. As an Azure-certified partner, we receive preferred pricing. However, AWS also has a very competitive solution. 

Ultimately, the best choice depends on your relationship with Microsoft.

Azure OpenAI doesn't use a traditional licensing model. Instead, it's interaction-based, meaning transactional. The cost depends on the complexity of the business use case and the amount of computing used within OpenAI.

It's important to engineer your solution carefully and implement controls. With any AI solution, there's a risk of operational expenses spiraling if the team doesn't put guardrails in place. 

Tools like Azure Synapse can help ensure usage stays within defined limits. This is true for any cloud technology – you need financial controls to prevent unexpected costs.

I would rate the pricing a five out of ten. It's reasonably priced for now. It will likely become more affordable over time. As more providers offer support services within the Microsoft ecosystem and as user feedback shapes the technology, we'll see improvements. Because the technology is only about 20 months old, there's a lot of potential for growth.

Definitely experiment with it. They're leaders in this space. For simple initial use cases, anyone with a basic understanding of IT operations and service management can get started. 

However, when scaling to an enterprise level with multiple use cases, it's essential to bring in experts. You need to consider security, financial risks, and potential reputational risks.

Remember, it's a powerful tool, and like any tool, how you use it matters. Ensure those using the tool fully understand how to do so responsibly.

Considering the current market and its competition, I'd rate it a nine out of ten. Considering them purely as a service provider in this line of business. It's still an evolving technology, so there's room to reach the full ten.  

We have many use cases for the solution, such as digitalizing records, a chatbot looking at records, and being able to use generative AI on them. 

Our customers are worried about data management, ethical, and security issues. 

I have been working with the solution for the last six months. 

The tool is stable. 

Azure OpenAI is scalable. 

The tool's deployment can take weeks to complete. 

I rate Azure OpenAI an eight out of ten. Don't focus on the technology, but on the problem you try to solve. 

The most valuable feature is the ALM. It can execute clarifications against security. 

The dialogue manager needs to be improved. 

I have been using Azure OpenAI for six months. 

It is a scalable solution. 

I would rate the overall solution a seven out of ten. 

When I write documentation, I use the solution for translating anything and researching.

Azure OpenAI is useful for benchmarking products. We can search for all metrics with AI tools and compare our products with other products.

The solution’s stability could be improved. Azure OpenAI should use more specific sources like academic articles because I can't find the source.

I have been using Azure OpenAI for more than a year.

I rate the solution a nine out of ten for stability.

I rate the solution’s scalability a seven out of ten.

The solution’s initial setup is very easy.

I can deploy the solution by myself in a few seconds.

We pay a licensing fee for Azure OpenAI. The solution's pricing is normal worldwide but expensive in Turkey because Turkey's currency is different.

Azure OpenAI provides me with a lot of benefits. It's quick and saves me time.

Overall, I rate the solution an eight out of ten.

We are currently exploring the solution's LLM chat and question-and-answer-based endpoints.

The most valuable feature of the solution is the accuracy of ChatGPT. It is something we cannot reproduce with any open-source LLM. Azure OpenAI is easy to use because the endpoints are created, and we just need to pass our parameters and info.

Azure OpenAI will be expensive if you want to implement it as a permanent solution for a customer. I have explored Azure OpenAI from a purely LLM perspective. Its endpoints are currently enough for us to communicate with their ChatGPT instance and get results. I don't know if Azure has anything implemented for images, videos, and other endpoints.

We have been exploring ChatGPT 3.5 and 4 versions for one and a half years.

Around 20 to 50 users were using Azure OpenAI for one of our projects.

Whenever we encounter issues, we try to raise them on the portal, and we get a resolution from there.

Microsoft takes care of the deployment. We just get our instance, and we have to communicate with it.

If you consider the long-term aspect of any project, Azure OpenAI is a costly solution. However, the solution is cheap if you just want to see results or try some POC in the initial stages. This is because you don't need to spin up your instance; you can just consume things and see the results.

Azure OpenAI is a straightforward solution. After configuring it, you will get your endpoints. You then need to call the endpoints and pass the details. Azure OpenAI is a straightforward tool that is implemented in such a way that even a fresher or junior developer can learn to use it easily.

Overall, I rate the solution an eight out of ten.